Add Gemma-3-270M-uk verbalizer (INT8 MatMul-only QInt8, 226MB) + license notice
Browse files
verbalizer/gemma-int8/GEMMA-NOTICE.txt
ADDED
|
@@ -0,0 +1,13 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
HolosTTS neural verbalizer — Ukrainian text normalization (numbers, dates, phones...).
|
| 2 |
+
|
| 3 |
+
Files:
|
| 4 |
+
model.onnx INT8 quantized ONNX (MatMul-only dynamic QInt8), 18 layers, KV-cache, vocab 38651
|
| 5 |
+
tokenizer.model SentencePiece (pruned), bos=2 start_of_turn=6 end_of_turn=7 eos=1
|
| 6 |
+
config.json model config (hidden 640, head_dim 256, sliding_window 6)
|
| 7 |
+
|
| 8 |
+
Derived from:
|
| 9 |
+
- skypro1111/gemma-3-270m-uk-verbalizer (license: gemma)
|
| 10 |
+
- Gemma 3 270M by Google LLC — subject to the Gemma Terms of Use
|
| 11 |
+
https://ai.google.dev/gemma/terms
|
| 12 |
+
|
| 13 |
+
Redistributed under the Gemma Terms of Use; use of these files is subject to those terms.
|
verbalizer/gemma-int8/config.json
ADDED
|
@@ -0,0 +1,57 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_sliding_window_pattern": 6,
|
| 3 |
+
"architectures": [
|
| 4 |
+
"Gemma3ForCausalLM"
|
| 5 |
+
],
|
| 6 |
+
"attention_bias": false,
|
| 7 |
+
"attention_dropout": 0.0,
|
| 8 |
+
"attn_logit_softcapping": null,
|
| 9 |
+
"bos_token_id": 2,
|
| 10 |
+
"dtype": "bfloat16",
|
| 11 |
+
"eos_token_id": [
|
| 12 |
+
1,
|
| 13 |
+
7
|
| 14 |
+
],
|
| 15 |
+
"final_logit_softcapping": null,
|
| 16 |
+
"head_dim": 256,
|
| 17 |
+
"hidden_activation": "gelu_pytorch_tanh",
|
| 18 |
+
"hidden_size": 640,
|
| 19 |
+
"initializer_range": 0.02,
|
| 20 |
+
"intermediate_size": 2048,
|
| 21 |
+
"layer_types": [
|
| 22 |
+
"sliding_attention",
|
| 23 |
+
"sliding_attention",
|
| 24 |
+
"sliding_attention",
|
| 25 |
+
"sliding_attention",
|
| 26 |
+
"sliding_attention",
|
| 27 |
+
"full_attention",
|
| 28 |
+
"sliding_attention",
|
| 29 |
+
"sliding_attention",
|
| 30 |
+
"sliding_attention",
|
| 31 |
+
"sliding_attention",
|
| 32 |
+
"sliding_attention",
|
| 33 |
+
"full_attention",
|
| 34 |
+
"sliding_attention",
|
| 35 |
+
"sliding_attention",
|
| 36 |
+
"sliding_attention",
|
| 37 |
+
"sliding_attention",
|
| 38 |
+
"sliding_attention",
|
| 39 |
+
"full_attention"
|
| 40 |
+
],
|
| 41 |
+
"max_position_embeddings": 32768,
|
| 42 |
+
"model_type": "gemma3_text",
|
| 43 |
+
"num_attention_heads": 4,
|
| 44 |
+
"num_hidden_layers": 18,
|
| 45 |
+
"num_key_value_heads": 1,
|
| 46 |
+
"pad_token_id": 0,
|
| 47 |
+
"query_pre_attn_scalar": 256,
|
| 48 |
+
"rms_norm_eps": 1e-06,
|
| 49 |
+
"rope_local_base_freq": 10000.0,
|
| 50 |
+
"rope_scaling": null,
|
| 51 |
+
"rope_theta": 1000000.0,
|
| 52 |
+
"sliding_window": 512,
|
| 53 |
+
"transformers_version": "4.57.6",
|
| 54 |
+
"use_bidirectional_attention": false,
|
| 55 |
+
"use_cache": true,
|
| 56 |
+
"vocab_size": 38651
|
| 57 |
+
}
|
verbalizer/gemma-int8/model.onnx
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:50a641a3caa87e70d3c035ab263fc32e7a6036315a59a4d69b72a140ac157f55
|
| 3 |
+
size 226461823
|
verbalizer/gemma-int8/tokenizer.model
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:cb21cc434439bff2d87bd521490a873a8c51e7c00606a415f1f90c6f3618e54b
|
| 3 |
+
size 725383
|